The New York State Bar Association has adopted Legal Ethics Guidelines for Lawyers' Use of Artificial Intelligence, which say what an hourly biller can and cannot charge a client for AI work

What was adopted, and what it is

On 31 October 2026 the House of Delegates adopted the report. The association's release describes it as guidelines to help attorneys use AI responsibly. The report's Section 4, which it labels the operative text, says the Guidelines "do not amend the NYRPC, create independent disciplinary obligations, or require a lawyer to use any particular technology."

That sets the weight. When this article says the report states a position, it means the Association's reading of existing rules, not a rule of court. One further detail: Section 4 describes itself as Guidelines "proposed for adoption", while the cover and the release say they were approved. We read that as drafting that was not updated, which is our inference.

What a lawyer can bill

Guideline 5A is the fee section, tied to NYRPC 1.5. For hourly billers, 5A(a) says "A lawyer who bills on an hourly basis may bill only for time actually expended, including time spent formulating inputs to an AI Tool and reviewing, verifying, and revising its outputs." It adds that the lawyer may not bill for time that would have been expended had an AI tool not been used, or bill for the value of the work product rather than the time spent.

On learning time, 5A(b) says "A lawyer may not bill a client for time spent acquiring general familiarity with an AI Tool." It allows time reasonably spent adapting or applying a tool to the particular matter, to the extent reasonable under NYRPC 1.5(a). The release puts this as no fees for learning how to use the tool. The report's line between general familiarity and matter-specific adaptation is the finer one.

For non-hourly fees, 5A(d) says efficiency gains do not, standing alone, require a reduction, but excessiveness is judged under all the factors in NYRPC 1.5(a). So the guideline treats hourly and flat fees differently.

Tool costs and client disclosure

On costs, 5A(c) says "The general cost of acquiring and maintaining AI Tools is ordinarily an overhead expense and may not be billed to a client as a disbursement" unless the charge is for services or expenses incurred for, or reasonably allocable to, the client's matter, the basis is disclosed under NYRPC 1.5(b), and the amount is reasonable. The release mentions reasonableness and disclosure; Guideline 5A(c) states those conditions expressly within the exception for client-specific charges.

Guideline 4A is the general disclosure section. Under it, the report lists situations in which disclosure to the client is required, including where the client asks how AI tools are used, where AI output is a significant basis for advice on a decision reserved to the client, and where the use affects the basis or rate of the fee or expenses. Outside those, 4A(c) says a lawyer may disclose AI use but is not required to, and routine functions such as spell checking need no disclosure.

Court submissions and supervision

The release says the report covers disclosing AI use in court submissions. The report's Guideline 8A is narrower than that suggests. It says lawyers make the inquiry required by NYRPC 3.1 and 3.3 and 22 NYCRR 130-1.1a before submitting a paper prepared with AI, and confirm every cited authority and quotation exists and supports the proposition. Whether a court requires disclosure or certification of AI use is, it says, governed by 22 NYCRR Part 161 and the rules of the individual court.

For supervisors, Guideline 9B says "A supervisory lawyer must not assume that a firm-approved tool or policy eliminates the need to review the work in the particular matter." Drafting principle 3.3.6 makes a similar point: organizational approval assists but does not replace matter-specific judgment. For a general counsel, the approved-tool list is a starting point, not a defense.

What we did not verify

We opened the NYSBA release and the full 106-page report PDF as extracted, and read the cover, executive summary, scope, the text of Guidelines 3A, 4A, 5A, 8A and 9B, and parts of 3B, 9A, 9C and 9D, plus the fee-opinion survey in Section 8.5. We did not read all 106 pages line by line, and we did not read the appendices, committee comments or the sections on bias, conflicts, agentic AI, unauthorized practice, pro bono, advertising or solicitation beyond their headings.

The file name of the report PDF indicates a version with comments. We cannot tell from the extract whether the House of Delegates adopted that exact text without amendment from the floor. The release says it adopted the report, and we rely on that.

We did not open ABA Journal coverage, 22 NYCRR Part 161 or the NYRPC. We refuse to claim that any New York court or disciplinary body has applied the Guidelines, or that a lawyer billing differently from them has breached a rule.

Related briefings

• New York Part 161 AI litigation rule
• NYC Bar Formal Opinion 2026-2 on AI recording of non-clients
• Texas Bar Opinion 705 on generative AI ethics
• North America AI regulation tracker
• AI Regulation News hub